Advanced weather forecasting for RES applications: Smart4RES developments towards high-resolution and Numerical Weather Prediction solutions to improve RES forecasting models
3 likes472 views
The ISGAN Academy webinar focused on the development of advanced weather forecasting techniques for renewable energy systems (RES), emphasizing the need for improved accuracy and reduced costs in forecasting. The presentation discussed various challenges in predicting renewable energy outputs, particularly from solar and wind sources, and proposed the use of high-resolution numerical weather prediction models to enhance forecasting capabilities. It highlighted the collaboration between science and industry to drive innovations that will support the future of renewable energy integration into smart grids.
![Further reading
[1] Gilbert, C., Messner, J. W., Pinson, P., Trombe, P. J., Verzijlbergh, R., van Dorp, P., & Jonker, H.
(2020). Statistical post-processing of turbulence-resolving weather forecasts for offshore wind
power forecasting. Wind Energy, 23(4), 884–897. https://doi.org/10.1002/we.2456
[2] Lindsay, N., Libois, Q., Badosa, J., Migan-Dubois, A., Bourdin, V. (2020). Errors in PV power
modelling due to the lack of spectral and angular details of solar irradiance inputs. Solar
Energy,197, 266-278 https://doi.org/10.1016/j.solener.2019.12.042
[3] Schalkwijk, J., Jonker, H. J. J., Siebesma, A. P., & Van Meijgaard, E. (2015). Weather
forecasting using GPU-based large-Eddy simulations. Bulletin of the American Meteorological
Society, 96(5), 715–723. https://doi.org/10.1175/BAMS-D-14-00114.1
[4] E.J. Wiegant, P. Baas, R.A.Verzijlbergh, B.Reijmerink, and S.Caires. The new frontier in
numerical metocean modelling: coupled high-resolution atmosphere wave interactions. In Wind
Europe Offshore, 2019.
[5] R.A.Verzijlbergh, H.J.J. Jonker, P. van Dorp, E.J. Wiegant, P. Baas, B.M. Meijer, and J. Coeling.
Breakthrough weather prediction technology enables wind turbine resolving resource assessments.
In Wind Europe, 2019.
35](https://image.slidesharecdn.com/isgansmart4resepisode3final3-210323161333/85/Advanced-weather-forecasting-for-RES-applications-Smart4RES-developments-towards-high-resolution-and-Numerical-Weather-Prediction-solutions-to-improve-RES-forecasting-models-35-320.jpg)
Advanced weather forecasting for RES applications: Smart4RES developments towards high-resolution and Numerical Weather Prediction solutions to improve RES forecasting models
- 1. ISGAN – International Smart Grid Action Network Matthias Lange, Quentin Libois, Remco Verzijlbergh 23 March 2021 ISGAN Academy webinar #27 Recorded webinars available at: https://www.iea-isgan.org/our-work/annex-8/
- 2. ISGAN in a Nutshell Created under the auspices of: the Implementing Agreement for a Co-operative Programme on Smart Grids 2 Strategic platform to support high-level government knowledge transfer and action for the accelerated development and deployment of smarter, cleaner electricity grids around the world International Smart Grid Action Network is the only global government-to- government forum on smart grids. an initiative of the Clean Energy Ministerial (CEM) Annexes Annex 1 Global Smart Grid Inventory Annex 2 Smart Grid Case Studies Annex 3 Benefit- Cost Analyses and Toolkits Annex 4 Synthesis of Insights for Decision Makers Annex 5 Smart Grid Internation al Research Facility Network Annex 6 Power T&D Systems Annex 7 Smart Grids Transitions Annex 8: ISGAN Academy on Smart Grids
- 4. Value proposition 4 ISGAN Conference presentations Policy briefs Technology briefs Technical papers Discussion papers Webinars Casebooks Workshops Broad international expert network Knowledge sharing, technical assistance, project coordination Global, regional & national policy support Strategic partnerships IEA, CEM, GSGF, Mission Innovation, etc. Visit our website: www.iea-isgan.org
- 5. Advanced weather forecasting for RES applications Smart4RES developments towards high-resolution and Numerical Weather Prediction solutions to improve RES forecasting models 23 March 2021 5 Matthias Lange, Quentin Libois, Remco Verzijlbergh
- 6. Agenda • Smart4RES in a nutshell • Challenges in forecasting of renewables • Towards Numerical Weather Prediction models dedicated to Renewables Energy Sources • High-resolution weather models 6 ADVANCED WEATHER FORECASTING FOR RES APPLICATIONS
- 7. Smart4RES in a nutshell ADVANCED WEATHER FORECASTING FOR RES APPLICATIONS 7
- 8. Smart4RES in a nutshell ADVANCED WEATHER FORECASTING FOR RES APPLICATIONS 8 • RES forecasting is a mature technology with operational tools and commercial services used by different actors • However, we want to make progress to improve the forecasting accuracy and to reduce costs of RES integration Smart4RES vision Science and industry closely co-operate to achieve outstanding improvements of RES forecasting by considering the whole model and value chain. Episode 3 Episode 2 Episode 4
- 9. Smart4RES webinar series 9 #1 June 2020 Introduction to Smart4RES - Data science for renewable energy prediction #2 Dec 2020 Extracting value from data sharing for RES forecasting #3 March & April 2021 #4 May 2021 #5 July 2021 Optimizing the value of storage in power systems and electricity markets #6 September 2021 Modelling tools for integrating RES forecasting in electrical grids Optimising participation of RES generation in electricity markets: new opportunities and the role of forecasting Season1: Towards a new Standard for the entire RES forecasting value chain Replay on Youtube Replay on Youtube Advanced weather forecasting for RES applications 3.1 NWP & High- resolution models 3.2 Data observations & assimilation
- 10. Smart4RES consortium ADVANCED WEATHER FORECASTING FOR RES APPLICATIONS 10 6 countries 12 partners 11/2019-4/2023 Funds: H2020 programme Budget: 4 Mio€ Duration: 3.5 years End-users Universities Research Industry Meteorologists Follow us! www.smart4res.eu
- 11. Challenges in forecasting of renewables 11
- 12. Challenges in forecasting of renewables • Solar power prediction: the fog situation 12 fog seen from satellite impact of fog on solar power forecast for Germany: individual weather models with large forecasting errors NWP based forecast corrected forecast production Depending on market situation and the trading strategy such a single forecasting error can be very expensive. For example: up to 0.25 Mio € in the most expensive hour and up to 0.75 Mio € over the day. source: energy-charts.info
- 13. Challenges in forecasting of renewables • Wind power prediction: fluctuations and ramps 13 measurement prediction State-of-the-art weather models provide a time resolution of 1 hour and do not predict fast changes
- 14. Towards Numerical Weather Prediction (NWP) models dedicated to Renewable Energy Sources (RES) 14
- 15. Innovative weather forecasts for RES – Challenges and Solutions • atmosphere is chaotic (initial state is critical) • NWP models have coarse resolution while physical processes occur at very small scales (clouds, rain, wind gusts, etc.) → Parameterizations needed Why is weather forecasting so difficult ? • using ensembles to explore all possible states • increasing spatial resolution to get rid of parameterizations • improving parameterizations What are the solutions ? Many physical processes occur at scales smaller than the model grid, and are not explicitly simulated. They are parameterized based on explicitly simulated large scale variables Schematic illustration of the use of ensembles in weather forecasts (UK Met Office)
- 16. Towards NWP models dedicated to RES Comparison between 1-hour (operational) and 5-min resolution outputs for wind speed forecasts (courtesy B. Alonzo) evaluating and calibrating NWP models accounting for RES scores and developing specific forecasters capabilities extracting additional relevant variables for RES (cloud optical thickness, direct/diffuse partition, spectral distribution of radiation, rapid wind fluctuations) providing higher temporal resolution outputs (NWP models have a ~1 min time step but standard output is 1 hour) Paths to improved RES forecasts Direct solar radiation in 6 spectral bands simulated by AROME around Toulouse (France)
- 17. High temporal resolution outputs • High frequency fluctuations in simulated wind speed are consistent with observations • Slight underestimation of these fast fluctuations in the ensemble mean but some members do capture them better • Autocorrelation demonstrates the forecast capability Low frequency variations and intra-hourly standard deviation of wind speed. Comparison between ensemble simulations and observations (courtesy B. Alonzo) Autocorrelation functions of low and high frequency wind fluctuations (courtesy B. Alonzo)
- 18. Ensembles (perturbed initial conditions, lateral forcing, surface or parameterizations) highlight how the atmospheric state can vary a lot from a member to another Different members can have very distinct behaviours Fog prediction (maps of liquid water content at level 1, 10 m high) around Paris for 6 members of an ensemble (courtesy T. Bergot) Paris airports The power of ensembles – fog Fog predictions (operational 12 members ensemble simulations) and observations (12 measurements) at CdG airport (Paris) in December 2018 (courtesy T. Bergot, R. Lestringant) LWC (g kg -1 )
- 19. • Ensembles are useful to estimate the uncertainty associated with a forecast • Uncertainties in RES relevant variables can be significant Relative standard deviation (standard deviation/mean) for wind (left) and global horizontal irradiance (right) across an ensemble of 25 members (courtesy B. Alonzo) The power of ensembles – wind, solar
- 20. The power of ensembles – Pseudo-deterministic simulations • Ensembles can be used to build pseudo-deterministic simulations (that look like deterministic simulations to the end- user) • These pseudo-deterministic simulations can be built in various ways, and can be very specific to the end-user needs • Pseudo-deterministic simulations can exceed ensemble mean forecasting performances Performances of various simulations with respect to observations. Here, the dedicated pseudo-deterministic simulation performs best (courtesy B. Alonzo) Definition of categories for wind speed, used to build pseudo- deterministic runs for wind power forecasts. Over each period, the most represented category across the ensemble is used.
- 21. Increasing spatial resolution • Fog formation is very sensitive to the state of the atmosphere close to the surface • Increasing vertical and horizontal resolution modifies turbulence and dynamics Higher vertical resolution better reproduces fog daily cycle near Paris airport (Philip et al., 2016) Differences in fog prediction for different spatial resolutions (2,5 km, 60 levels vs 1 km, 156 levels ; Ragon, 2020) Fog cover (%) Time along one night
- 22. High-resolution weather models Large Eddy Simulation (LES): the future 22
- 23. Operational LES model of the HornsRev offshore wind farm https://vimeo.com/whiffle/hornsrev Gilbert, C., Messner, J. W., Pinson, P., Trombe, P., Verzijlbergh, R., Dorp, P. Van, & Jonker, H. (2019). Statistical Post- processing of Turbulence-resolving Weather Forecasts for Offshore Wind Power Forecasting. Wind Energy, 1–16. https://doi.org/10.1002/we.2456
- 24. Resolved motions Sub-grid motions Large Eddy Simulation Model grid in physical space showing schematically the turbulent motions (eddies) In spectral space the energy of different eddies peaks around a certain size (small wavenumber k are the largest eddies) A turbulent spectrum: energy as function of wavenumber Fourier transform dx = grid size dx The energy spectrum of turbulent flow in Large Eddy Simulation (LES)
- 25. Flow dependent Flow independent LES resolves all relevant physical processes and only parametrizes homogeneous turbulence Wavenumber k Energy The energy spectrum of turbulent flow in Large Eddy Simulation (LES)
- 26. High-resolution allows to resolve the surface in detail Higher resolution and more explicit modelling of: • Forest and vegetation resolved in more detail (canopy parameterization) • Explicitly model obstacles like buildings and wind turbines Tiggelen, M. Van. (2018). Towards improving the land- surface- atmosphere coupling in the Dutch Atmospheric Large- Eddy Simulation model (DALES). MSc Thesis, Delft University of Technology. Land use Terrain and buildings Turbines 15 km Surface scheme controls momentum, heat and moisture exchange
- 27. Aerial & satellite images Land use and vegetation properties Obstacles, terrain and orography (mountains) Turbine types and locations Large-scale weather forecast GPU-Resident Atmospheric Simulation Platform (GRASP) Operational forecasting with LES Static boundary conditions A schematic representation of different inputs to a the LES forecasting model
- 28. Parametrization: expressing the sub-grid processes in terms of resolved quantities See for example https://www.ecmwf.int/en/elibrary/18714-ifs-documentation-cy45r1-part-iv-physical-processes resolved resolved resolved parametrized parametrized Typical processes that are parameterized in NWP: • Turbulence • Large-scale clouds • Convective clouds • Surface drag • Radiation • Precipitation • Surface energy balance Schematic view on transport by sub-grid processes Physics parameterizations in NWP
- 29. Parametrization: expressing the sub-grid processes in terms of resolved quantities See for example https://www.ecmwf.int/en/elibrary/18714-ifs-documentation-cy45r1-part-iv-physical-processes resolved parametrized Typical processes that are parameterized in LES: • Turbulence • Large-scale clouds • Convective clouds • Surface drag • Radiation • Precipitation • Surface energy balance (in high resolution) The LES grid is fine enough to resolve turbulence, clouds and the surface. “Assume less, compute more” Explicit modelling of: • Wind turbines • Canopies • Buildings • Turbulence • Clouds / fog Physics parameterizations in LES
- 30. Traditional weather forecast Satellite image Whiffle forecast Country scale LES already possible on multi-GPU systems
- 31. Smart4RES use case: Rhodes Important forecast characteristics for smarter grid management: • Fast fluctuations in wind power • Spatio-temporal patterns • Good forecast accuracy on complex terrain A nested LES forecast is produced • Outer domain captures Rhodes ~ 100m resolution • 4 inner domains on wind farms: ~ 40m resolution • Figure shows variability in individual wind turbine forecasts • Compare with smooth signal of large-scale weather model (era5)
- 32. Planned innovations in Smart4RES Local observations • Clouds • Wind • Radiation • Pressure • Power • SCADA wind • SCADA temp • … Static boundary conditions Large-scale weather model data GPU-Resident Atmospheric Simulation Platform (GRASP) New in Smart4RES: assimilation of local observations in LES Data assimilation : estimating the state of a system (here: the atmospheric model) given a set of observations and the model dynamics minimize obj = f (modeled state – observed state) subject to model dynamics
- 33. The future of numerical weather prediction Wind farm level EU level Global Forecast domain Time 2030 2019 Country level LES: scaling up the domain size NWP: increasing the resolution 2030 2019 Grid cell size Next generation weather model: - Turbulence and cloud resolving ( = LES ! ) - Uses big data and massive computational power - Supports energy transition and climate adaptation Synoptic scale Cloud system Convection Turbulence Who will be first?
- 34. Take-away messages • Numerical models contain much more than operational outputs • The natural evolution of NWP is to increase spatial resolution → Now at a stage called grey-zone turbulence in between parameterizations and LES (~ 500 m resolution) → LES might be the future of NWP • Ensembles are becoming the norm to explore uncertainties • Increasing amount of observations, including non-conventional, are likely to improve forecasts via data assimilation Attend our next webinar! → for very short-term, direct observations remain very powerful because models give a statistical view of the atmosphere, not perfectly punctual in space and time → combination of various inputs is promising, in particular with increased use of artificial intelligence Follow us! www.smart4res.eu Evolution of Météo-France computation power
- 35. Further reading [1] Gilbert, C., Messner, J. W., Pinson, P., Trombe, P. J., Verzijlbergh, R., van Dorp, P., & Jonker, H. (2020). Statistical post-processing of turbulence-resolving weather forecasts for offshore wind power forecasting. Wind Energy, 23(4), 884–897. https://doi.org/10.1002/we.2456 [2] Lindsay, N., Libois, Q., Badosa, J., Migan-Dubois, A., Bourdin, V. (2020). Errors in PV power modelling due to the lack of spectral and angular details of solar irradiance inputs. Solar Energy,197, 266-278 https://doi.org/10.1016/j.solener.2019.12.042 [3] Schalkwijk, J., Jonker, H. J. J., Siebesma, A. P., & Van Meijgaard, E. (2015). Weather forecasting using GPU-based large-Eddy simulations. Bulletin of the American Meteorological Society, 96(5), 715–723. https://doi.org/10.1175/BAMS-D-14-00114.1 [4] E.J. Wiegant, P. Baas, R.A.Verzijlbergh, B.Reijmerink, and S.Caires. The new frontier in numerical metocean modelling: coupled high-resolution atmosphere wave interactions. In Wind Europe Offshore, 2019. [5] R.A.Verzijlbergh, H.J.J. Jonker, P. van Dorp, E.J. Wiegant, P. Baas, B.M. Meijer, and J. Coeling. Breakthrough weather prediction technology enables wind turbine resolving resource assessments. In Wind Europe, 2019. 35
- 36. Smart4RES webinar series 36 #1 June 2020 Introduction to Smart4RES - Data science for renewable energy prediction #2 Dec 2020 Extracting value from data sharing for RES forecasting #3 March & April 2021 #4 May 2021 #5 July 2021 Optimizing the value of storage in power systems and electricity markets #6 September 2021 Modelling tools for integrating RES forecasting in electrical grids Optimising participation of RES generation in electricity markets: new opportunities and the role of forecasting Advanced weather forecasting for RES applications Season1: Towards a new Standard for the entire RES forecasting value chain Replay on Youtube Replay on Youtube 3.1 NWP & High- resolution models 3.2 Data observations & assimilation
- 38. Back-up slides